EL PUEBLO DE ISRAEL

This Part argues against using either only the plaintiff’s work or the defendant’s work as the baseline for assessing substantiality. The quantitative, percentage-style baselines invoked in cases such as Harper &

Row and Newton v. Diamond needlessly ignore important context—either

the nature of the use, or the nature of the original work.257 Indeed, outside of the context of linear, literary works, the concept of quantitative measure becomes incoherent.

For similar reasons, courts should not ignore important dissimilarities between the original and accused works by focusing solely on the plaintiff’s work. The tendency of some courts to disregard important dissimilarities results from a misunderstanding of the Sheldon rule and its misapplication far beyond its original purpose. Instead, courts should take a primarily qualitative approach to substantiality, considering relative importance contextually, using both works as baselines and not dismissing dissimilarities as irrelevant to the infringement analysis.

The current approach to the baseline and dissimilarity problems—to the extent there is any consensus at all—is embodied in the Sheldon rule. Recall that Sheldon instructs courts to ignore the changes that an alleged infringer has made because “no plagiarist can excuse the wrong by showing how much of his work he did not pirate.”258 Sheldon leads courts to take the plaintiff’s work as the baseline for comparison, which has the effect of making the original work the more prominent, and biases the

257. See supra Part II.B (reviewing case law examples of the baseline problem). 258. See supra notes 176–79 and accompanying text.

726 WASHINGTON UNIVERSITY LAW REVIEW [VOL.93:681

comparison in favor of the plaintiff.259 (This concern is exacerbated when the plaintiff’s copyrighted work is well known, a not uncommon scenario in copyright disputes.260) Many courts invoke Sheldon to justify using the plaintiff’s work as the baseline when measuring the quantitative significance of the defendant’s copying.261 Some courts have extended the

Sheldon rule to disregard dissimilarities between the two works and

changes that the defendant has made.262

The Sheldon rule is triply wrong. First, the “tacking on” concern that motivated Sheldon is overstated, and courts have not been careful to limit the Sheldon rule to its original context of linear, literary works. Second, a strict application of Sheldon causes courts to ignore the context of the defendant’s use, context that may be critical to understanding whether the alleged infringement should be actionable. Sheldon is thus unfair to defendants when it ignores the new material that they added and the context in which the alleged infringement was made. Finally, Sheldon is incorrect to focus myopically on the plaintiff’s work as the baseline. This can result in unfairness to the plaintiff, such as in the Antonick case, when larger works are effectively penalized for involving additional expression. Courts’ tendency to shift baselines in an ad hoc fashion may well be an indication of the tensions created by a strict application of Sheldon.

The concern that motivated Sheldon is the “tacking on” problem—the idea that an infringer could escape liability simply by adding a mass of new, extraneous material to his infringing work.263 This concern is overstated. It is not apparent in the facts of Sheldon itself, and it is far from clear that the tacking-on problem is a realistic scenario. In any event, even presuming the concern is real, there are better and more direct ways to deal with the problem of “manipulating” the de minimis inquiry.264

Blatant attempts to avoid infringement by tacking on additional material can be easily ferreted out directly by courts, and dismissed via equitable doctrines.265

259. See supra notes 13, 108–09 and accompanying text (summarizing results from cognitive science that greater similarity is perceived when the more prominent of two concepts is treated as the baseline).

260. See, e.g., Warner Bros. Entm’t Inc. v. RDR Books, 575 F. Supp. 2d 513, 518 (S.D.N.Y. 2008) (finding infringement of Harry Potter series of novels); Castle Rock Entm’t, Inc. v. Carol Publ’g Grp., Inc., 150 F.3d 132, 135 (2d Cir. 1998) (finding infringement of Seinfeld television series).

261. See supra notes 164–75 and accompanying text (discussing Newton v. Diamond). 262. See supra notes 209–24 and accompanying text (discussing Tufenkian). 263. See supra notes 179–80 and accompanying text.

264. Cf. 3 PATRY, supra note 5, § 9:65 (“[Sheldon] has rightly been cited as authority against attempts to manipulate the de minimis [infringement] inquiry.”).

2016] REFRAMING SIMILARITY ANALYSIS IN COPYRIGHT 727

In the more usual case, what was added by the defendant is not “tacked on,” but rather important, and should not be ignored. The addition of new expression by the defendant is typically not extraneous, unrelated material—as envisioned by Sheldon—but instead precisely the type of expression one might want to encourage through copyright. Take Newton

v. Diamond as an example.266 Though one could view the rapping and music that the Beastie Boys added as “tacked on,” that is a ridiculous claim—the material added is the heart of how the Beastie Boys used the sample to make a new work. Notably, the Sheldon rule is precisely the opposite of what courts do in the context of fair use, where much of the emphasis is rightly on whether the defendant transformed the original work and added “something new.”267

In the context of similarity, the defendant’s new material is relevant both to the perceived similarity of a lay observer, as well as to any economic harm caused by the infringement.268

It is therefore odd to dismiss the defendant’s new material or dissimilarities between the two works out of hand. Why would we not consider the defendant’s actual use in the context of his new work? In

Newton, the Beastie Boys should certainly prevail, but not because their

looping is “irrelevant.”269

Of course it’s relevant. If “Pass the Mic” were nothing but Newton’s flute sample looped for three minutes—no rapping, no added samples and instrumental music—would it not be a different, and tougher, case? Moreover, although ignoring the defendant’s use in

Newton worked to the advantage of the defendant, usually it is used in the

opposite manner, as it was in Sheldon itself. This is true in the Tufenkian case, for example, where Sheldon was invoked to prevent the defendant from relying on dissimilarities to negate the infringement claim.270 In other

words, although Sheldon saves the Beastie Boys from having to explain looping, it leaves them unable to rely on the fact that they added new samples, rapping, and live instruments over the Newton sample. Similarly,

that equitable doctrine of estoppel is available in copyright actions); Saxon v. Blann, 968 F.2d 676, 680 (8th Cir. 1992) (confirming that the doctrine of unclean hands is applicable in copyright actions).

266. See supra notes 164–75 and accompanying text (discussing Newton). 267. Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569, 579 (1994).

268. See Jeanne C. Fromer & Mark A. Lemley, The Audience in Intellectual Property

Infringement, 112 MICH.L.REV.1251, 1255 (2014) (arguing that infringement in intellectual property

regimes should only be found when works are similar from a technical perspective, and cause market harm from the view of consumers); Lape, supra note 10, at 202 (arguing that substantial similarity requires a finding of “economic harm” or other “injury that may have an impact on authors’ incentive to create”).

269. Newton v. Diamond, 388 F.3d 1189, 1195 (9th Cir. 2003).

728 WASHINGTON UNIVERSITY LAW REVIEW [VOL.93:681

it leaves the defendants in cases like Tufenkian or Adler unable to rely upon the important dissimilarities between their works and the original works. Considering only what the defendant copied, without considering how the defendant used what was copied in context, or the changes she made, is a form of willful blindness that tends to bias the inquiry against defendants.

The Sheldon rule can sometimes be unfair to plaintiffs as well. In particular, using the plaintiff’s work as the measure for quantitative substantiality effectively penalizes the plaintiff for creating more expression. Consider the facts of the Antonick case, for example.271 There, no substantial similarity was found because the original work was complex—a football simulation video game—even though EA had copied Antonick’s code for plays and formations. When the original work happens to be large—be it Ford’s ponderous memoir in Harper & Row, or the millions of lines of code in Antonick or Oracle—percentage-style measures using the plaintiff’s work as the baseline dilute the significance of what was taken. In Harper & Row, three hundred words is a trivial percentage of Ford’s memoir, but it might still be substantial in context. As Table 2 shows, percentage-based measures can vary widely depending on the baseline.

TABLE 2:QUANTITATIVE MEASURES OF HOW MUCH WAS COPIED USING

VARIOUS BASELINES

Baseline Harper & Row Newton v. Diamond Oracle v. Google Antonick v. Electronic Arts Plaintiff’s work 0.1% (“insubstantial”) 2% of running time; greater % of score 3% of copied APIs; trivial % of Java small Defendant’s work

“at least 13%” looped background element; significant (1/5) part of song trivial % of Android small Absolute taking 300 words, plus paraphrasing

six seconds; three note phrase 7,000 lines of code code for plays and formations

2016] REFRAMING SIMILARITY ANALYSIS IN COPYRIGHT 729

Using the plaintiff’s work as the baseline thus seems perverse if copyright is designed to incent new expression. Taken literally, it effectively penalizes original creators for creating more expression. The real issue in these cases is whether what was taken was substantial in

itself, and qualitatively significant in the context of both the original and

accused works. The court in Antonick should consider whether the plays and formation code is significant in overall qualitative importance to the games, how many lines of code total were copied, and how much of that was commonplace or dictated by functionality. To dismiss the plays and formation code as insignificant only because it is small piece of a larger game misses the point. In this respect, Justice O’Connor’s focus on the qualitative import of the expression in Harper & Row was correct. However, the sleight of hand involved in invoking the defendant’s work as the baseline in Harper & Row—“at least 13% of the infringing article”—is misleading.272 These percentage-style measures are red herrings. Would

Harper & Row really be a different case if Ford’s memoir were only

10,000 words, or if The Nation’s article were twice as long? Does it really matter in Oracle how many millions of lines of code are in Java or Android? Probably not, and courts should not imply otherwise.

Furthermore, the notion of a percentage measure is unsound for non- linear works. In literary works, we can at least count the number of words or lines of code taken—though, of course, paraphrasing may present some difficulties. For musical works, one instinct is to look to the percentage of the running time, as the Newton court did. But it is not clear why that measure is superior to using the number of notes, the percentage of the score, or the prominence of a sound in the mix. Running time also becomes deceptive when there is more than one sample or instrument playing simultaneously, as is usual. In theory, one could try to apportion this—say, 25% of two minutes for the first sample, etc.—but this quickly becomes arbitrary. More importantly, such quantitative measures only avoid the real issues of qualitative significance. In Newton, the courts should be asking whether the idea of singing into a flute is even protectable expression at all, not whether “six seconds” or “three notes” is a significant part of Newton’s composition. If Newton had squeezed a Charlie Parker-style flurry of notes into the six-second sample, or if “Choir” was only twenty seconds long, the measures the majority relied upon in Newton would greatly change. However, the fundamentals of the case would not be truly different.

730 WASHINGTON UNIVERSITY LAW REVIEW [VOL.93:681

The difficulties of quantitative measurement are even more serious for pictorial, graphic, and sculptural works.273 For visual works intended to be perceived as an entirety, one can scarcely begin to imagine how to define a percentage appropriation. Take the two bubble-blowing Santa ornaments in Adler as an example.274 How does one calculate what “percentage” of the ornaments are the same, and how much of that was due to protectable elements? We are at a loss to say. In theory, perhaps, one could analyze a two-dimensional work pixel-by-pixel and count the pixels that are identical. But this method is coherent, and then just barely, only in the special case of pastiche.275 A pixel-by-pixel analysis would likely yield zero or a trivial number of similarities in a case where an artist’s visual style was copied—Steinberg v. Columbia Pictures Industries is a well- known example—even though it is clear to a lay observer that something was taken (whether we think that something is protectable or not).276 This pixel-based measure also runs aground without a way to exclude all unprotected material. The rugs in Tufenkian, for example, might be highly similar on pixel percentage basis, but that is likely only because they are both derived from the same public domain sources.

In sum, courts should focus mainly on the contextual, qualitative importance of what was copied in assessing its significance in substantial similarity. If the Sheldon rule must be invoked at all, it should be limited to its original purpose—cases where a plaintiff or defendant attempts to manipulate the infringement inquiry by adding extraneous material. To the extent courts must consider the copying quantitatively at all, the best course may be to measure the expression copied as an absolute matter. Such a measure looks to the total number of words copied, the number of lines of code, or the length of a piece of music, but always in light of the qualitative significance of that expression in context, including its level of originality and its economic significance.277 The absolute measure at least gets courts thinking about the right questions—for example, whether what was copied was too short, too commonplace, or too functional to be

273. See 17 U.S.C. § 102(a)(5) (2014).

274. See supra notes 225–31 and accompanying text (discussing Adler).

275. See Cariou v. Prince, 714 F.3d 694, 705–12 (2d Cir. 2013) (assessing whether series of appropriation art pastiche of photographs from earlier works was a fair use).

276. 663 F. Supp. 706 (S.D.N.Y. 1987). That case involved a poster for the movie “Moscow on the Hudson” that evoked the style of Saul Steinberg, an artist for the New Yorker, and in particular his famous work “View of the World from 9th Avenue.” Id. at 708–11.

277. Cf. Busek, supra note 10, at 1795–1803 (arguing for an approach to substantial similarity based on the degree of possible expressive variation); Lape, supra note 10, at 194–206 (arguing for an approach to substantial similarity based on substantial economic or other harm to the plaintiff).

2016] REFRAMING SIMILARITY ANALYSIS IN COPYRIGHT 731

protectable expression—and not its relative shortness vis-à-vis the plaintiff’s or defendant’s work.278

Similarly, courts should not disregard material added by the defendant, or important dissimilarities that undermine perceived similarity. Ignoring dissimilarities is another unintended result of Sheldon’s dictum. It is a mistake to adhere to this rule literally, as the Second Circuit did in Tufenkian. The balance struck by Judge Newman in Warner Bros. is a sounder approach.

CONCLUSION

This Article has sought to bring attention to the framing problems that plague courts’ analyses of similarity in copyright infringement. Courts take a distressingly inconsistent approach to the issues of baselines, timing, perspective, and dissimilarities when structuring the comparison between the original and the accused works. The result is widespread confusion, unpredictability, and contradictory results. This Article argues that courts should take a flexible, contextual approach to similarity’s framing problems. Such an approach is fairer to both parties, more predictable, and truer to copyright’s purpose of promoting the creation and dissemination of knowledge. It avoids the manipulation and biases created by the more rigid approaches that predominate in the case law. Most importantly, it serves to guarantee substantial similarity’s role as a robust and vital limitation on the scope of copyright.

278. This is one of the few things that the Federal Circuit got right in Oracle. The fact that “7,000 lines of code” were copied is probably more relevant than the fact that those lines are only a trivial aspect of Java or Android. See Oracle Am., Inc. v. Google Inc., 750 F.3d 1339, 1356 (Fed. Cir. 2014). However, the qualitative significance of the code in that case makes it obvious that no infringement should be found. The key issue in that case was not the total amount taken, but its standard, commonplace, and functional nature. Arguably, Google might be permitted to copy the whole of the APIs.